
REV. A
–7–
ADXL105
Output Scaling
The acceleration output (A
OUT
) of the ADXL105 is nominally
250 mV/
g
. This scale factor may not be appropriate for all appli-
cations. The UCA may be used to increase the scale factor. The
simplest implementation would be as shown in Figure 14a.
Since the 0
g
offset of the ADXL105 is 2.5 V
±
625 mV, using a
gain of greater than 4 could result in having the UCA output at
0 V or 5 V at 0
g
. The solution is to add R3 and VR1, as shown
in Figure 14b, turning the UCA into a summing amplifier. VR1
is adjusted such that the UCA output is V
DD
/2 at 0
g
.
C
R1
OUT
R2
IN
f
–3dB
=
1
2
p
CR1
GAIN = –R2
V
MID
a. 1-Pole Low-Pass Filter
0.22
m
F
OUT
20k
V
IN
f
–3dB
= 30Hz
20k
V
0.18
m
F
V
MID
b. 2-Pole Bessel Low-Pass Filter
R1
OUT
R2
IN
f
–3dB
=
1
2
p
CR2
GAIN = –R2
C
R3
R3 ~
V
MID
V
MID
c. 1-Pole High-Pass Filter
44.2k
V
OUT
IN
f
–3dB
= 10Hz
59k
V
0.39
m
F
0.39
m
F
V
MID
d. 2-Pole Bessel High-Pass Filter
Figure 15. UCA Used as Active Filters*
Device Bandwidth vs. Resolution
In general the bandwidth selected will determine the noise floor
and hence, the measurement resolution (smallest detectable
acceleration) of the ADXL105. Since the noise of the ADXL105
has the characteristic of white Gaussian noise that contributes
equally at all frequencies, the noise amplitude may be reduced
by simply reducing the bandwidth. So the typical noise of the
ADXL105 is:
Noise
(
rms
) = (225
μ
g
/
√
Hz
)
× (√
Bandwidth
×
K
)
Where
K
≈
1.6 for a single-pole filter
K
≈
1.4 for a 2-pole filter
So given a bandwidth of 1000 Hz, the typical rms noise floor of
an ADLX105 will be:
Noise = (225
μ
g
/
√
Hz
)
×
(
√
1000
×
1.6)
= 9 m
g
rms for a single-pole filter
and
Noise = (225
μ
g
/
√
Hz
)
×
(
√
1000
×
1.4)
= 8.4 m
g
rms for 2-pole filter
Often the peak value of the noise is desired. Peak-to-peak noise
can only be estimated by statistical means. Table I may be used
for estimating the probabilities of exceeding various peak values
given the rms value. The peak-to-peak noise value will give the
best estimate of the uncertainty in a single measurement.
Table I. Estimation of Peak-to-Peak Noise
Nominal Peak-to-
Peak Value
2
×
rms
3
×
rms
4
×
rms
5
×
rms
6
×
rms
7
×
rms
8
×
rms
% of Time that Noise Will
Exceed Peak-to-Peak Value
32%
13%
4.6%
1.2%
0.27%
0.047%
0.0063%
The UCA may be configured to act as an active filter with gain
and 0
g
offset control as shown in Figure 16.
0.1
m
F
OUT
IN
GAIN = 2
f
–3dB
= 30Hz
0.1
m
F
100k
V
47k
V
47k
V
10k
V
V
DD
47k
V
Figure 16. UCA Configured as an Active Low-Pass Filter
with Gain and Offset
EMC and Electrical Noise
The design of the ADXL105 is such that EMI or magnetic
fields do not normally affect it. Since the ADXL105 is ratiomet-
ric, conducted electrical noise on V
DD
does affect the output.
This is particularly true for noise at the ADXL105’s internal
clock frequency (200 kHz) and its odd harmonics. So maintain-
ing a clean supply voltage is key in preserving the low noise and
high resolution properties of the ADXL105.
One way to ensure that V
DD
contains no high frequency noise is
to add an R-C low-pass filter near the V
DD
pin as shown in
Figure 17. Using the component values shown in Figure 17,
noise at 200 kHz is attenuated by approximately –23 dB. As-
suming the ADXL105 consumes 2 mA, there will be a 100 mV
drop across R1. This can be neglected simply by using the
ADXL105’s V
DD
as the A-to-D converter’s reference voltage as
shown in Figure 17.
*For other corner frequencies, consult an active filter handbook.